Page 1 of 2
BSE stop
Posted: Thu Oct 14, 2021 12:49 am
by sdwang
Dear developers,
I am running BSE for a 2D system, and the BSE stopped after the Kernel, and seems the diagonal process has a problem. The system is metal for PBE while it is a semiconductor under GW approximation, and I used the GW energe for the BSE. Does Yamob have a problem with BSE for the metal?(the version of Yambo for other system is OK for me)
Thanks!
Shudong
Re: BSE stop
Posted: Thu Oct 14, 2021 7:56 am
by Daniele Varsano
Dear Shudong,
the situation you describe should not have any problem. Impossible to say more without looking at the report file.
Best,
Daniele
Re: BSE stop
Posted: Tue Nov 16, 2021 1:48 pm
by pyadav
Dear Yambo team,
I'm running BSE calculation for a material (which is a DSM) but it is getting terminated repeatedly. I would really appreciate any help.
PFA.
Thanks,
Pushpendra
Re: BSE stop
Posted: Tue Nov 16, 2021 2:47 pm
by Daniele Varsano
Dear Pushendra,
can you please post at least one of the log file in order to see where the code stop?
and importantly please check if you have some error message in one of the log files:
e,g,
Code: Select all
grep -i ERR l_* in the LOG directoty
Best,
Daniele
Re: BSE stop
Posted: Tue Nov 16, 2021 5:26 pm
by pyadav
Dear Daniele,
Thanks for the response.
I'm not able to see any error message in the log files. there are almost 120 log files (corresponding to each processor) but attaching two of them from different trials. PFA.
bse_termination.zip
Wishes,
Pushpendra
Re: BSE stop
Posted: Tue Nov 16, 2021 5:36 pm
by Daniele Varsano
It is not easy to spot the problem, I'm just wondering if it could be a memory issue.
Can you try to run a calculation reducing the bands in BSE ? e.g.
Code: Select all
% BSEBands
45 | 56 | # [BSK] Bands range
%
if this is the case then it would be useful to track the used memory by enabling the flag:
in the configure and recompiling the code
Best,
Daniele
Re: BSE stop
Posted: Tue Nov 16, 2021 5:46 pm
by pyadav
Dear Daniele,
I'll try reducing the BSE bands first. I will see if it works.
Thank you for the suggestions
Regards,
Pushpendra
Re: BSE stop
Posted: Thu Nov 25, 2021 12:37 pm
by pyadav
Dear team Yambo,
I'm running a BSE calculation for a metallic system and getting it terminated with the following error message
in the report file
Code: Select all
[ERROR] STOP signal received while in[03.03.03] Main kernel loop
[ERROR]EMPTY WF index @ b47 k4 spin1 ID0
also seeing
Code: Select all
[03] Bethe Salpeter Equation @q1
================================
[WARNING] Null BSE kernel block dimension(s) found.
[WARNING] Empty workload for CPU 1
Attaching the report, log, and input files for reference. PFA.
Seeking help to figure out the issue.
Thanks,
Pushpendra
Re: BSE stop
Posted: Thu Nov 25, 2021 12:44 pm
by Daniele Varsano
Dear Pushendra,
this happens because your system is metallic:
Code: Select all
[X] Filled Bands : 44
[X] Metallic Bands : 45 54
[X] Empty Bands : 55 400
you have partial occupation from band 45 to 54.
You need to include at least one fully empty and one fully occupied state in the BSE Hamiltonian while your setting is:
Code: Select all
% BSEBands
49 | 58 | # [BSK] Bands range
%
and not fully occupied bands are found.
Best,
Daniele
Re: BSE stop
Posted: Fri Nov 26, 2021 8:05 am
by pyadav
Dear Daniele,
Yes, it runs now. But in that case, since there are 10 partially filled bands between completely empty and completely filled bands so I'm able to take only 1 completely empty and 1 completely filled band in BSE calculation because after this any increment in the band numbers (suppose 2 VB +10 Mettalic Bands+ 2 CB =14 bands) breaks the BSE calculations probably due to memory issue.
I do not know why there is no transition from the fully filled bands to partially filled bands or partially filled bands to empty bands. I'm not sure if there is some selection rule that does not allow these transitions or it is code specific case. These are the open questions coming to my mind, I will appreciate it if you could shed some light on them.
Also is there any tag in Yambo to include the Drude term (the intraband term)?
Thank you for the help.
Best
Pushpendra.